intel/fs/cse: Split create_copy_instr into three cases

Previously, we tried to combine all cases where the instruction being
CSE'd writes to more than one MOV worth of registers into one case with
a bit of special casing for LOAD_PAYLOAD.  This commit splits things so
that LOAD_PAYLOAD is entirely it's own case.  This makes tweaking the
LOAD_PAYLOAD case simpler in the next commit.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

diff --git a/src/intel/compiler/brw_fs_cse.cpp b/src/intel/compiler/brw_fs_cse.cpp
index 9b8978040324f8a5e45f852acaa28a343d99af35..bd917baaa6d83bd45e325638404d137e94dbe907 100644 (file)
--- a/src/intel/compiler/brw_fs_cse.cpp
+++ b/src/intel/compiler/brw_fs_cse.cpp
@@ -207,30 +207,30 @@ create_copy_instr(const fs_builder &bld, fs_inst *inst, fs_reg src, bool negate)
       DIV_ROUND_UP(inst->dst.component_size(inst->exec_size), REG_SIZE);
    fs_inst *copy;
 
-   if (inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD ||
-       written != dst_width) {
-      fs_reg *payload;
-      int sources, header_size;
-      if (inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD) {
-         sources = inst->sources;
-         header_size = inst->header_size;
-      } else {
-         assert(written % dst_width == 0);
-         sources = written / dst_width;
-         header_size = 0;
-      }
-
+   if (inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD) {
       assert(src.file == VGRF);
-      payload = ralloc_array(bld.shader->mem_ctx, fs_reg, sources);
-      for (int i = 0; i < header_size; i++) {
+      fs_reg *payload = ralloc_array(bld.shader->mem_ctx, fs_reg,
+                                     inst->sources);
+      for (int i = 0; i < inst->header_size; i++) {
          payload[i] = src;
          src.offset += REG_SIZE;
       }
-      for (int i = header_size; i < sources; i++) {
+      for (int i = inst->header_size; i < inst->sources; i++) {
+         payload[i] = src;
+         src = offset(src, bld, 1);
+      }
+      copy = bld.LOAD_PAYLOAD(inst->dst, payload, inst->sources,
+                              inst->header_size);
+   } else if (written != dst_width) {
+      assert(src.file == VGRF);
+      assert(written % dst_width == 0);
+      const int sources = written / dst_width;
+      fs_reg *payload = ralloc_array(bld.shader->mem_ctx, fs_reg, sources);
+      for (int i = 0; i < sources; i++) {
          payload[i] = src;
          src = offset(src, bld, 1);
       }
-      copy = bld.LOAD_PAYLOAD(inst->dst, payload, sources, header_size);
+      copy = bld.LOAD_PAYLOAD(inst->dst, payload, sources, 0);
    } else {
       copy = bld.MOV(inst->dst, src);
       copy->group = inst->group;